This paper investigates the possible relationship between the most important start-up motive of the founder and subsequent performance of the firm, measured three years after the start-up. Using a large longitudinal database with 7000+ cases, the analysis was performed in two steps. First, the effect of the start-up motives on firm survival was investigated, while further analysis was done to study whether the performance of those firms that were still active were related to the start-up motives. In this study, start-up motives did not show any predictive power on survival. In relation to subsequent performance, the two motives “need to work independently” and, to a lesser extent, “unemployment” came out positively correlated and significant, although effect sizes were very small.

It is intuitively appealing to expect the start-up reason of the entrepreneur to influence firm performance. In the newly formed firm, strategic scope and resource acquisition is closely linked to the entrepreneur and it is in these initial stages where the relationship between entrepreneurial drive and firm performance should reveal itself at its strongest. However, previous empirical findings have been rather dismissive of this intuitive logic. In the present study, five pre-defined start-up reasons assessed during the firms’ first year in operation (plus an open ended alternative) are related to performance in terms of number of employees, sales and perceived profitability.

Many studies have investigated the start-up motives of business founders. Through an extensive literature review, Scheinberg and MacMillan (1988) derived 38 different motives for starting a business. In their own empirical study with the telling title “An 11 country study of motivations to start a business” they investigated the stated motivations of a sample of 1 402 owner-managers. By analyzing their empirical data using principal component analysis, the original set of 38 items was reduced to 21 and grouped into 6 uncorrelated components or dimensions. These were labeled 1. Need for Approval, 2. Perceived Instrumentality of Wealth, 3. Degree of Communitarianism, 4. Need for Personal Development, 5. Need for Independence and 6. Need for Escape. Birley and Westhead (1994), later used the reduced set of motives as a basis for their study of 405 principal owner-managers of new independent business in Great Britain. The purpose of this study was twofold. First, the authors wanted to investigate whether there existed any differences in the motives of owner-managers to start a business. Secondly, if there indeed were any differences, would there be any differential impact on subsequent performance? Their analysis was carried out in three steps. The first step involved grouping 22 items into components using principal component analysis. In this case, the number of components was set to 7, of which 5 were in conceptual agreement with the 6-component matrix presented by Scheinberg and MacMillan (1988). The second step involved using the components as clustering variables in a cluster analysis. The retained solution yielded a total of 6 clusters that was subsequently tested for possible performance differences. However, the analysis of performance revealed no significant differences between clusters in relation to sales and employment levels.

In this paper we build on these previous studies in an analysis of the relationship between start-up motivations and subsequent outcomes. Using a large, longitudinal database we will first relate start-up motives to survival of the business three years later. Among the firms that are still active we will also relate start-up motives to three different performance measures.

The Sample

The sample used in this paper consists of 7 256 new enterprises that were surveyed in a joint effort by the Swedish National Board for Industrial and Technical Development (NUTEK) and Statistics Sweden. The sample has been surveyed twice, measuring initial conditions in the first wave and performance in the second (see Figure 1).

In more detail, the sample was created in the following way. In February 1995, a sampling frame was constructed by Statistics Sweden, covering all legal forms of business activities registered during 1994. To be considered a registered business activity (or enterprise) in this sample, the business founders do not have to make any formal registration with a business register. Reporting VAT or income from a business activity in the personal income statement is sufficient. Using cross-referencing across four different registers, it was found that 74 600 new business registrations were made in Sweden during 1994. From this first sampling frame, 14 500 businesses in agriculture, forestry, hunting, fishery and real estate were excluded. Another 2 700 businesses in various industries were also excluded since it could be clearly established from register information that they were take-overs. Thus, the final sampling frame consisted of 57 400 newly registered enterprises, from which a proportional stratified sample of 14 000 was drawn. Strata were constructed according to industry, legal form and geographical location (municipality).

In the first wave, a mail questionnaire was sent out to the businesses in the sample, yielding a response rate of 86%. This unusually high figure for a mail survey was in part due to the fact that the core questions were part of a compulsory business survey. Also, extensive use of telephone interviews in case of non-response in the mail survey contributed considerably to the final results. Out of the approximately 12 000 responses, 7 256 fell into the definition of genuinely new enterprises and these were later selected for the follow-up. According to the definition provided by NUTEK and Statistics Sweden, a genuinely new enterprise has to be a new business activity in a new (independent) legal entity. It cannot be a take-over or a mere re-registration of an already existing business (e.g. change of legal form). In Sweden, no less than 48% of all new business registrations are not new business activities, but simply existing business activities put into a new legal entity. This emphasizes the importance of some sort of screening when addressing research questions related to new business activities using survey methodology.

In August 1998, a second wave of questionnaires was issued to all of the 7 256 enterprises that were recognized as genuinely new enterprises in the initial survey. A similar method of inquiry was used, but there was a heavier reliance on telephone interviews to minimize non-responses. Out of the 7 256 businesses that were selected for the second wave, responses were obtained from 6 377, resulting in a response rate of 87.9%. However, in 97 cases the original business had been sold or merged with another firm. Although these firms were obviously surviving at the time of sale, they were excluded from further analysis since it would be difficult to assess the performance of the original activity, both conceptually and empirically.

Businesses that were non-response cases in the second mail survey became the objects of intense investigation to try to contact individuals connected to these businesses. Since some of the questions in the follow-up were compulsory by law, simple refusal to answer was extremely rare. The final non-response cases are those businesses that could not be contacted by any means (telephone or mail). Since access to register information on the non-response businesses was not an issue, a common characteristic of the final non-responses was a protected telephone number or no listing. We conclude that would be very difficult to run a legitimate business under such circumstances and that it is likely that these businesses have suspended operations, be it temporarily or definitely. The non-response cases in the follow-up survey are therefore considered as “not active” and therefore excluded from the analysis. The method used to produce this sample has the advantage of excluding businesses that are not new. Since the data set originally was designed for inferential statistics, it is also representative of the population of genuinely new businesses that was started in Sweden in 1994.

Measurement of Motives

The method of measuring start-up motives used in the present study is rather straightforward. In the first wave, respondents were asked about their most important motive for starting their business. Five alternatives were predefined, complemented by an open-ended alternative. The six categories were re-coded into five dichotomous variables. Table 1 shows the alternatives together with frequencies for valid responses.

This method of measuring start-up motives is markedly different from that used in the cited studies in two important respects. First, the cited studies used scales, our does not. Secondly, the cited studies allowed for multiple motives, our does not. Thus, the measurement we have at our disposal is rather crude in comparison. However, there are some redeeming features of the measurements used in this study that makes them defensible. First of all, the classification of motives used in this survey has proven quite robust over time. The origins of this classification can been traced to an (unpublished) pilot study made by Statistics Sweden in 1990, on behalf of the Swedish Government. Testing some 30 different motives, the final set only contained four fixed alternatives and one open-ended. This set of motives has since then been in use for the annual survey on business start-ups in Sweden, from which the sample use in this study is derived. In 1992, a fixed alternative relating to unemployment was added (because of its increased importance in the then prevailing deep recession) but otherwise the original classification has been in use for 10 years without alteration. The open-ended alternative typically receives 5% of total number of valid responses. Unfortunately, the responses from the open-ended responses have not been available to us for reasons of anonymity. However, officials at Statistics Sweden state that the themes detected show a great deal of variety. There is at least anecdotal evidence that a common theme is dissatisfaction with previous work situation due to e.g. need for more flexible working hours or even allergies.

To be comparable to the previous studies, it would desirable to have our fixed alternatives corresponding to each of the basic components arrived at in the previous studies. We will use the components or dimension that was originally proposed by Scheinberg and MacMillan (1988) as a gauge since they are elaborated on the basis of an international sample and therefore could be expected to be more robust across different empirical settings. The motives “to realize my ideas,” “unemployment or risk thereof,” “to make money” and “to work independently” all seem rather straightforward to classify. The motive “Needed on the market” is a somewhat different animal relating to opportunity and is ambiguous in relation to the proposed classification. Scheinberg and MacMillan’s instrument does not capture this type of externally attributed start-up reasons. The classification of the motives is indicated in the right column of Table 1. The reader may have noted that some components are not captured by the five motives. There are, however, reasonable causes why the components “Degree of Communitarianism” and “Need for Approval” would not make it into the top 5 motives for starting a business in Sweden. To begin with, Scheinberg and MacMillan (1988) found Sweden to be particularly low on “Degree of Communitarianism.” Secondly, being a business owner is (or was) not a high status occupation in Sweden. Rather, they are traditionally met with some suspicion. Therefore, it is reasonable to expect a low fraction of the population to the start-up of a business as a vehicle for social approval. Hence, we would not expect the component “Need for Approval” to score particularly high in this specific empirical setting.

Control Variables

In the results section we will present regression results both with and without control variables. The control variables added in the second regression model are dichotomous except for initial size, which is the sales estimate for the first year of operation coded into six classes: less than SEK 30’, SEK 30’–99’, SEK 100’–299’, SEK 300’–499’, SEK 500’–999’, SEK 1000’ and above (SEK 1000 = $116, April 2000). The sex of the founder is a self-reported measure in three categories: male, female and joint leadership. In the regression models, this item is re-coded into two dichotomous variables Male and Female. The variable Start-up experience indicates whether the founder has started at least one business prior to the sampled firm. Parallel start-up reports whether the person answering the questionnaire has started another business since 1994. The variable Incorporated indicates that the firm was incorporated at start-up. This could be interpreted in two ways. Either we may see it as an indicator of initial financial resources since an incorporated company (aktiebolag) has to have an initial equity capital of at least SEK 100 000. Another possible way is to interpret it as an indicator of commitment. However, we prefer the former interpretation since it makes less assumptions about the characteristics of the founder or the founding team. The variable Start-up course refers to whether or not the founder has received any formal training specifically aimed at preparing applicants for the business start-up situation. These courses typically range from one day to two weeks. The variable Immigrant pertains to people who were born outside Sweden or has at least one parent born outside of Sweden. The concept of immigrants in a Swedish context mirrors that of ethnic minorities in the U.S. indicating marginality as well as reduced access to resources. The last control variable is Start-up grant. This is a financial support to registered unemployed who want to start their own business. The start-up grant is equivalent to the applicants pending unemployment benefits and is normally received for six months with possible extension a further six month for groups that are considered underprivileged, especially in rural areas. The grant and the unemployment benefits are mutually exclusive in Sweden.

Dependent Variables

Using the entire original sample we use active versus not active three years after start-up as our first dependent variable. Among firms that are still active there are three additional dependent variables to consider: employment level, sales and perceived performance. The employment level is measured in terms of full-time equivalents including the owner. Both employment level and sales have been transformed using log10 since the distributions were heavily skewed towards the lower-end. The third measure, perceived profit, is an index created from two scaled items. The first item measures how the profit level of the business is perceived (“not satisfactory” through “very satisfactory” and the second item relates to a statement on how well the business provides for the founder and his or her family (“not enough” through “very well”).

Active vs. Not Active Three Years After Start-up

This initial analysis was carried out by means of logistic regression, because of the dichotomous nature of the active/not active dependent variable. Two models were used. The first specified only the start-up motives as explanatory variables while the second model also included the control variables. These models are identical to the ones listed in the linear regression models presented in Table 3. Neither model succeeded in predicting survival any better than just assuming that all the firms survived, see Table 2. These results are in line with some previous efforts in terms of predicting firm “survival” based on factors observable at the time of the start-up (Cooper et al. 1994; Dahlqvist et al. 2000). The predictive value has been marginal at best, depending on the precision and detail of the information available. The logistic regression analysis was also carried out for each of nine industries defined as the single digit level of industries in the industrial classification in NACE rev. 1 and results are consistent.

Performance Among Still Active Firms

The linear regression models were specified in the same way as for the logistic regression analysis. Only firms classified as “active” and “active, but in a new legal unit” were included, in all 4 091 cases. The results are presented in Table 3. The pattern is rather consistent over the various measures of the dependent variable. Since the sample is large, even very small effects become significant at the conventional 5% risk level.. Therefore we have opted to mark only values with a significance level of 1% or better. Since the sample constitutes a sizable part of the population, the p-values are conservative.

One pattern across the different independent variables is the significant effects from the motives “unemployment or risk thereof” and “need to work independently” in the first model. The motive “needed on the market” has a weak effect in the first model, but this goes away when the control variables are included in model 2. No motive is significant at the one per cent level in model 2 when tested in relation to sales. With respect to the both objective measures of performance, it could be said that effect size is halved when the control variables are included concurrent with an increase in the overall adjusted R² by a factor of ten to twenty. In other words, the first two motives are significant and stable over the various measures of the independent variable in regression model 1, but the adjusted R² is verging on the trivial with around three per cent of the variance explained. Sub-analyses of nine industry sectors confirm this conclusion. In regressions without controls the first two motives consistently get positive and significant effects, at least in service industries. In the regression with controls, no start-up motive is consistently ascribed a significant effect across industries.

There are some caveats about the interpretation of these results. First of all, the introduction of first year sales as a control variable could be regarded as an alteration of the dependent variable from size to growth. In the first model, the independent variables are expected to influence for conditions at the start-up as well as subsequent development. A second condition to point out is that initial size is measured in terms of sales.. That is why this variable achieves a high beta when regressed on sales in 1997. Also the overall R² is relatively high in this case at 0.549.

In the case of “unemployment or risk thereof” some of the “edge” is taken off this motive by the fact that a large part of those starting with this motive also received a start-up grant. As touched upon briefly in the introduction to the variables, a start-up grant is given to only to people that are registered as unemployed. The applicants are eligible for a six-month program in which they will receive the monthly equivalent of their unemployment benefits. Because we do not know which people was de facto unemployed prior to their start-up, we cannot distinguish between the assumed disadvantages of unemployment and the effect of perceiving that you started your business for unemployment reasons.

In this paper we have attempted to replicate Birley and Westhead (1994) to the extent possible. However, our data permitted us only to explore the possible link between start-up reasons and performance. The studies differ in some important respect, notably on the use of scales and possibility to account for multiple motives. The characteristics of the data set at our disposal are that it is rich on cases but rather economical on measurements. Still, we think the results have a fair amount of validity. First of all, most of the motives listed in our survey belong to different dimensions put forth in earlier studies. Moreover, these measures have proved to be very robust over time, at least in this specific empirical setting.

In order to see if the start-up motive has some predictive value on survival we carried out logistic regression analyses. In the face of the classification tables produced, the conclusion can only be that is not possible to predict survival based on start-up motives alone. Not even the model including several control variables succeed in predicting survival much better than assuming that all firms survived. Notwithstanding the empirical problems of measuring constructs, there are certain theoretical complications in devising models for prediction of survival based on factors observable at start-up.

One problem pertains to the difference between entry and exit cost, which is difficult to account for in the type of model used in this paper. To give but one example, initial capital is typically expected to increase performance and hence survival by helping the firm to overcome entry costs. But the initial capital endowment is beside the point when discussing exit costs, which is an essential factor to consider in relation to exit. In this case, the type and tradability of the firm’s assets would have to be considered.

A second problem is that the persistence of the small firm is not solely a question of “the survival of the fittest,” although this is an underlying assumption in models of firm survival that are based on initial conditions. Rather, the existence of the small firm is highly influenced by factors outside the firm’s competitive environment. At any given time, decisions to persevere or to abandon is balanced against the outside options of the owner-founder(s) and the value of these alternatives will be based estimates of future states rather than historical values. If the alternative options for the founders are not included in our models, it is only natural to expect unexplained variation when examining the relative performance of firms that are discontinued. (For a thorough discussion on exit, see Gimeno et al. 1997.)

The conclusion is that models for understanding survival need to take into account the value of the founder’s outside options at any specific time. It is therefore not very surprising that the type of exercises that we have partly engaged in here have not proven very successful in terms of predictive value of business survival (Cooper et al. 1994; Dahlqvist et al. 2000). Since the choice to terminate a business that is tied to a specific business owner would include a multitude of options outside the firm, filtered through cognitive processes, it is really not that surprising that models of firm survival from relatively simple measurements at start-up leaves the majority of the variance unexplained.

Turning to the performance of the firms that were still active, Birley and Westhead (1994) used clusters as a basis to test performance. They found no significant differences between clusters. Our results on the other hand, show significant and positive betas for “unemployment” and “need for independence” in the first regression model. When the control variables are added, however, the betas for the start-up motives drop radically and only “need for independence” remains significant in relation to the objective measurements of performance. This pattern is reasonably stable over industries, although statistically significant betas are mostly achieved in the service sector. The results suggest that there indeed is some impact of start-up reasons on subsequent performance among firms that are still active, but that this impact is very weak to say the least. An interesting result is that both “unemployment “ and “need for independence” both are significant in the second regression model when regressed on the perceived profitability. This result may in part be explained by differences in cognition but as we already have alluded to in relation to firm survival, it is reasonable to expect the availability of outside options to have some impact, at least for the unemployment motive.

CONTACT: Jonas Dahlqvist, Jönköping International Business School, P.O. Box 1026, SE-551 11 Jönköping, Sweden; (T) 46 36 157547; (F) 46 36 16 10 69; jonas.dahlqvist@jibs.hj.se

Birley, S. and P. Westhead. (1994) “A taxonomy of Business Start-up Reasons and Their Impact On Firm Growth and Size.” Journal of Business Venturing 9: 7–31.

Cooper, A.C., F.J. Gimeno-Gascon and C.Y. Woo. (1994) “Initial Human and Financial Capital as Predictors of New Venture Performance.” Journal of Business Venturing 9: 371–395.

Dahlqvist, J., P. Davidsson and J. Wiklund. (2000) “Initial Conditions as Predictors of New Venture Performance: A Replication and Extension of the Cooper et al. Study.” Enterprise & Innovation Management Studies 1: 1–18.

Gimeno, J., Folta, T.B., Cooper, A.C. and Woo, C.Y. (1997) “Survival of the Fittest? Entrepreneurial Human Capital and the Persistence of Underperforming Firms.” Administrative Science Quarterly 42: 750–783.

Scheinberg, S. and I.C. MacMillan. (1988) “An 11 Country Study of Motivations to Start A Business.” Frontiers of Entrepreneurship Research. Wellesley, MA: Babson College.